1. Dimensional Data Design - Data Mart Life Cycle
1.
Introduction
A data mart is a persistent physical store of operational and aggregated data statistically processed data
that supports businesspeople in making decisions based primarily on analyses of past activities and
results. A data mart contains a predefined subset of enterprise data organized for rapid analysis and
reporting.
This section provides the life-cycle steps for developing and enhancing data marts.
Additional standards that may be relevant to data mart projects include the following:

Data Administration (DA) standards, which are available on the CMS Internet;

Data Base Administration (DBA) Standards, which are available on the on the CMS
Internet.

Systems Development Standards. CMS's system development standards are outlined in the
CMS Integrated IT Investment & System Life Cycle Framework, which is available on the
CMS Intranet.
The Dimensional Data Design- Data Mart Life Cycle diagram depicts the milestones, control points, and
deliverables as they occur during the following steps:
Dimensional Data Design Processes:
Initiate Data Administration Services
Data Mart Development
Data Mart Enhancement
Supplemental Information:
Appendix: Dimensional Data Modeling Standards
Appendix: Related DA Standards, Forms, Tools
Key Deliverable(s):
 Dimensional Logical Data Model
 Dimensional Physical Data Model
 Data Definition Language
 Production Data Mart
Exhibit 1. Dimensional Data Design- Data Mart Life Cycle diagram
Dimensional Data Design - Data Mart Life Cycle
Initiate Data
Administration Services
Define Data Mart
Requirements
Enhanced / Changed
requirements
Verify
Verify
verified
adjustment
needs
Create / Enhance Logical
View of Dimensional Data
Model
Tuning / Error correction
Verify and Approve
verified
adjustment
needs
Develop Physical Data
Model
Verify and Approve
verified
adjustment
needs
Create Development Data
Mart
Test and Verify
Perform Acceptance
Testing
Verify
Development
Enhancement
Implementation
Verify
Maintenance
phase
2.
Initiate Data Administration Services
Introduction:
It is important to request DA support early in the Requirements phase of a data mart project. Early
involvement enables the Project Data Analyst to gain a thorough understanding of project needs and to
plan the data mart appropriately. Use the following procedure to initiate a data services request.

Submit a Data Services Request
Deliverable(s):
 A Data Services Request
1.2.1 Submit a Data Services Request
Project Manager
A. Early in the Requirements phase of the project, the
Project Manager/GTL makes a request for data support
and services. To provide background for this request, the
Project Manager completes the required sections of the
Data Services Request and submits the form according
to its instructions. The Data Service Request procedure
and form is available from the main Data Administration
web page.
Data Services Manager
B. Evaluate the data services request. If it is appropriate
and feasible to provide DA support, assign a qualified
Project Data Analyst to the project, appropriately to the
level of anticipated complexity.
Project Data Analyst
C. Meet with the project team. Team members introduce
themselves and explain their roles. The Project Data
Analyst provides an overview of DA standards and the
data mart development life cycle.
From this point on, the Project Data Analyst is a member
of the project team and is included in team meetings and
communications.
3.
Data Mart Development
Introduction:
The objective of data mart design is easy access to relevant information. The star schema is the design
method of choice at CMS, as it enables a relational database to emulate the analytical functionality of a
multidimensional database.
The following processes describe the prescribed activities and standards for development of data marts at
CMS, using the star schema approach.

Define Data Mart Requirements

Create Logical View of Dimensional Model

Approve Logical View of Dimensional Model

Develop Physical View of Model

Approve Physical View of Model

Create and Test Development Data Mart

Perform Acceptance Testing

Implement in Production
Deliverable(s):
 Data Mart Requirements
 Approved Dimensional Logical Data Model
 Approved Dimensional Physical Data Model
 Data Definition Language
 Production Data Mart
1.3.1 Define Data Mart Requirements
Business Analyst
Project Data Analyst
A. Identify the business requirements for the new data mart.
The requirements specification should include, but is not
limited to, the following information:







Categories of users and the likely business questions to
be asked by each group;
Geographic location(s) of users;
Data required to answer user questions;
Source(s) for the data;
Likely volumes and access patterns;
Architectural concept views of the data mart as it may
relate to enterprise business, data, applications and
technology architectures;
Security requirements.
Project Data Analyst
B. Work with the Business Analyst to identify and provide
models and metadata for the data source(s) to be used by
the data mart. An Enterprise Data Model (EDM) extract
including the source entities and attributes may be
provided, as well as individual logical and physical
models of source databases.
C. (Future) The enterprise Conformed Dimensions and
Conformed Facts will be examined for suitability, and
any required changes or additions will be submitted to
Central DA and the Enterprise Architecture team.
Business Analyst
Project Data Analyst
D. Conduct a review of documented data requirements. The
requirements review provides an opportunity for the
Business Analyst to confirm that the requirements
specification is complete and accurate. If a review is
held, members of the project team may attend.
Project Manager,
Business Owner/Sponsor
E. Sign-off on documented data requirements.
1.3.2 Create Logical View of Dimensional Model
Data Administrator
Project Data Analyst
A. Provide source models and metadata for enterprise
Conformed Dimensions and Conformed Facts.
("Conformed" dimension attributes and facts are those
designed for standardized use by data marts across the
agency.)
Data Administration provides source models and
associated metadata for these data structures.
Project Data Analyst
B. Work with the Business Analyst and subject matter
experts to develop the logical view of the dimensional
model, based on the approved documented data
requirements. See the Dimensional Model Standards
for a complete description of this deliverable.
The model must also be compliant in the naming of
Entities and Attributes, including use of standard
business terms. The Standard and Abbreviations Terms
List is available from the Standards Terms page
(accessible from the Data Administration home web
page). If a needed term is not on the list, follow the
procedure outlined on the Standard Terms page.
All CMS data models must be documented in the CMS
standard data modeling tool. For more information about
use of the standard modeling tool, see the DA Standard
Tools topic, which is accessible from the Data
Administration home web page.
1.3.3 Approve Logical View of Dimensional Model
Project Manager A. Every dimensional data mart project must conduct a review
of the logical view of the data model. This review should be
completed prior to the development of physical names.
If the model has been previously approved or enhancements
are being made to an existing approved dimensional model,
only the changes to that model are reviewed. If the existing
model was never reviewed, or if a new model was created,
the entire model must be reviewed.
Review Purpose
The purpose of the model review is to: verify that the
dimensional model meets user requirements; confirm that
preferred data sources have been identified; validate the
technical quality of the model; and ensure that existing data
structures and metadata are being reused appropriately.
Review Timing
It is advisable for the Project Data Analyst to hold interim
reviews during the development of the dimensional model.
Interim reviews provide the opportunity to familiarize the
Business Analyst and development team with the model and
obtain feedback on business and technical modeling issues.
If an interim review is held, a representative from Data
Administration should participate.
The final review of the dimensional model takes place when
the Project Data Analyst believes that the model meets all
known business requirements, meets DA standards, and
includes all required metadata.
Review Package Contents
Provide the necessary information to guide participants
through the review process. The review package must
include the following:

A soft copy of the of the data model. Distribute this via
email.

A separate printed diagram for each fact table and its
associated dimension tables. (Creating a separate subject
area in the model for each fact table and its associated
dimension tables will enable printing of separate
diagrams.)

An entity/definition report, including all entities (tables).

An entity/attribute/attribute-definition report, including
all attributes (columns).

A report including valid values and value descriptions for
any code sets documented for this data mart. See the
Dimensional Model Standards for information on code
sets (domains) that must be documented.

Business requirements documentation. Include an
overview of the business requirements. Include user
views (reports or screen designs) if available. Reviewers
may request additional documentation e.g., detailed
business requirements and, notes from interviews with
users.
Project Data Analyst B. Schedule reviews of the dimensional model to include three
components: the review kick-off meeting, the syntax or
technical review, and the review against business
requirements. At the end of the review, reviewers are
expected to approve the model by email message.
Project Manager, C. Arrange and conduct review kick-off meeting; prepare and
distribute review package.
Project Data Analyst
Project Manager, D. Participate in kick-off meeting.
Project Data Analyst,
Project Database
Administrator,
Data Administrator
Project Data Analyst E. Conduct the review session(s) and performs the tasks below.

Distribute the review package.

Provide information on project status and critical dates.

Walk through model at a high level (informational), pointing
out the major fact and dimension tables.

Establish responsibilities, events, and dates for the remainder
of the review.
Business Analyst F. Resolve issues requiring business needs.
Project Data Analyst G. Review Model against business requirements. Structure and
conduct walkthrough sessions. Explain the way in which the
model addresses the business requirements. Document and
problems with solutions or alternatives.
Business Analyst, Project H. Confirm that the model meets business requirements and
Data Analyst
communicate the outcome – whether it is concurrence or
disagreement. Point out any inconsistencies between model
and business requirements. Raise other issues as applicable.
Project Database I. Provide guidance for the anticipated data mart DBMS
Administrator
deployment.
Project Data Analyst J. Conduct the actual review session and perform the tasks
below.

Walk through the model in an orderly manner,
pointing out the major fact and dimension tables.

Identify data sources and explains the major processes
used to derive data from these sources.

Explain expected usage and navigation patterns and
the way in which the model will support these.

Explain any unusual features of the model. These
must be documented in a cover sheet, as stated in the
Dimensional Model Standards.
Project Manager, Business K. Approve model after all required changes have been made
Owner / Sponsor,
Data Administrator
Project Data Analyst L. Coordinate the transmission of approvals by email message.
When all reviewers are ready to sign off, Data
Administration sends them an email message asking for
confirmation of their approval by return email. The Project
Data Analyst forwards these approvals to Data
Administration; the project team and all reviewers are to
receive a copy of the approval messages.
M. Package the dimensional model with approval messages.
Project Data Analyst
Pass the approved model package to Data Administration,
where it is saved in the DA model library.
1.3.4 Develop Physical View of Dimensional Model
Project Data Analyst A. This phase includes development of the first-cut physical
view and final physical view of the model.
Generate a first-cut physical view in the standard modeling
tool. Participate in end-of-phase review.
Data Administrator, B. Assist in generating first-cut physical view as needed.
Review/approve physical names. Participate in end-of-phase
Database Administration
review.
Project Data Analyst C. With the assistance of Data Administration, generate the
first-cut physical view in the standard modeling tool. The
first-cut physical view includes physical names for the
logical elements. Detailed standards and procedures for
generating physical names may be found in Assign a Column
or Element Name.
Note: All elements in the physical model, including
physical-only elements, must have meaningful definitions. If
a single file with logical and physical views is used for the
logical and physical models, all definitions will be available
in both models automatically. If separate files are used, the
modeler must ensure that the definitions are brought over
into the physical model.
1.3.5 Approve Physical View of Dimensional Model
Project Data Analyst, A. Review the physical names with Data Administration. Data
Administration approves the names or recommends changes
Data Administrator
and, when appropriate, a follow-up review.
Data Administration has no role in the development of the
physical view after this point, except to assist in developing
physical names for any additional attributes. For information
on updates to logical and physical names, see Assign a
Column or Element Name.
Data Administration participates in the end-of-phase review
to ensure that any new elements comply with naming
standards, and that meaningful definitions have been
provided for all elements.
Data Administrator B. When the physical names have been approved, pass a copy
of the model and any associated documentation to the Project
Database Administrator. See the Database Administration
web page for more information on DBA standards and
procedures for the target DBMS environment.
Project Data Analyst, C. Obtain the dimensional model with sign-offs indicating
approval of the logical view by the Business Owner /
Project Manager
Sponsor, Data Administrator, and Project Manager. Database
Administration approval may also be required.
Forward the approved model and approval messages to Data
Administration.
Data Administrator D. After the end-of-phase review, the approved dimensional
model is saved in the DA model library.
1.3.6 Create and Test Development Data Mart
Database Administrator
A. Use the approved physical data model and generate
DBMS Data Definition Language (DDL) for the data
mart.
Database Administrator
B. Deploy the Dimensional Data Mart to the development
DBMS environment.
Notify the Project Manager of the test environment
availability.
1.3.7 Perform Acceptance Testing
Project Manager
A. Notify the Business Owner / Sponsor and business users
(or designated testers) of the available data mart test
environment.
Business Users,
Business Analyst
B. Test the data mart in the Validation/Acceptance Test
environment and verify that it meets business needs.
Project Database
Administrator
C. Support the data mart test environment as needed.
Project Data Analyst,
Project Database
Administrator
D. Participate in end-of-phase review. If there are changes
to the data mart requirements, enter modifications in the
appropriate dimensional model -- logical or physical
data model. Changes to models must be resubmitted
to the approval processes at the highest level of change.
For example, if the logical model is changed, restart the
process at Approve Logical View of Dimensional
Model; if the physical model is changed, restart the
process at Approve Physical View of Dimensional
Model. To expedite the process, limit the review and
approvals to the changed areas.
1.3.8 Implement in Production
Database Administrator
Database Administrator
A. Prepare production operational facilities and
components and deploy the data mart to the production
DBMS environment.
B. Monitor and support the data mart in the production
DBMS environment.
4.
Data Mart Enhancement
Introduction
All enhancements to a dimensional data mart, whether major or minor, must be incorporated in the
dimensional model for the data mart and comply with the Dimensional Modeling Standard. All changes
must be reported to Data Administration (DA) and Database Administration (DBA). In addition, major
enhancements -- such as those addressing new high-level business requirements, supporting new business
processes, and/or requiring additional tables--must follow the life cycle steps described in this document.
Enhancements to data marts begin with a baseline of the existing operating data mart. The processes to
meet new data mart requirements are:

Enhance Data Mart Requirements

Update Existing Logical View of Dimensional Model
Note: the remaining processes consists of steps that are the same as those performed for development of a
Data Mart.

Approve Logical View of Dimensional Model

Develop Physical View of Model

Approve Physical View of Model

Create and Test Development Data Mart

Perform Acceptance Testing

Implement in Production
Deliverable(s):
 New Data Mart Requirements
 Approved Dimensional Logical Data Model
 Approved Dimensional Physical Data Model
 Data Definition Language
 Production Data Mart
1.4.1 Enhance Data Mart Requirements
Data Administrator A. Provide extracts from the Enterprise Data Model (EDM)
and/or provides models and associated metadata from the
source databases that will feed the data mart.
Business Analyst B. Identify the new business requirements for the data mart.
Project Data Analyst
The requirements specification should include, but is not
limited to, the following information:

Any new categories of users and the likely business
questions to be asked by each

Geographic location(s) of new users

Additional business needs of existing users

Data required to address new business needs

Source(s) for the data

Any changes to volumes and access patterns

Any changes to security requirements
Identify new business requirements, data to support
requirements, and sources for data.
Project Data Analyst C. Work with the Business Analyst to identify and provide
models and metadata for the data source(s) to be used by the
data mart. An Enterprise Data Model (EDM) extract
including the source entities and attributes may be provided,
as well as individual logical and physical models of source
databases.
Business Analyst D. Conduct a review of documented data requirements. The
requirements review provides an opportunity to confirm that
Project Data Analyst
the requirements specification is complete and accurate. If a
review is held, any member of the project team may attend.
1.4.2 Update Existing Logical View of Dimensional Model
Data Administrator A. If the approved dimensional model for the existing data mart
is stored in the DA model library, provide a copy for the
project team.
Provide source models and metadata for enterprise
Conformed Dimensions and Conformed Facts.
("Conformed" dimension attributes and facts are those
designed for standardized use by data marts across the
agency.)
Project Data Analyst B. The dimensional model documents both the business
metadata (business names and definitions) and technical
metadata (the physical star schema structure) for the data
mart. The approved model for the existing data mart should
be available from Data Administration. This model must be
reviewed and if necessary updated to ensure that it accurately
represents the existing data mart.
Develop the logical view of the dimensional model. See the
Dimensional Data Modeling Standard for a complete
description of this deliverable.
Work with the Business Analyst develop and develop the
logical view of the dimensional model in the CMS standard
modeling tool.
Business Analyst C. Provide additional information on business requirements as
needed.
Project Data Analyst D. Review the existing dimensional model and if necessary
update it. Ensure that the model accurately reflects the
existing data mart.

Enhance the logical view of the dimensional model in
CMS’s standard modeling tool.

If any new "conformed" dimension tables are included in
the model, Data Administration provides source models
and associated metadata for these tables, if available.
("Conformed" dimension tables are those designed for
use by multiple data marts.)
Project Data Analyst E. If the approved dimensional model for the existing data mart
is stored in the DA model library, Data Administration
provides a copy for the project team. Note: If the
dimensional model for the existing data mart is not in the DA
model library, the project team is responsible for obtaining
the existing model or creating a new model. All models must
comply with the Dimensional Data Modeling Standard.
Update the dimensional model as needed to ensure that it is
synchronized with the existing operating data mart. This upto-date baseline dimensional model then becomes the
starting point for modeling the enhancements.
Project Data Analyst F. Follow the remaining approval and testing processes
necessary for data mart implementation. See Data Mart
Enhancement.
5.
Dimensional Data Modeling Standards
Dimensional Model Definition
Diagramming and Documentation Conventions
Naming Conventions
Relationships
Required Logical Metadata
Best Practices
1.
Dimensional Model Definition
A dimensional model is a graphical representation of information structures that allows business
information to be viewed from many perspectives. Dimensional data design is applicable in an analytical
processing or decision support environment. Dimensional modeling captures metadata similar to that in
an Entity-Relationship (E-R) model but uses a different structure--i.e., the star schema. The primary
components of the star schema are fact tables and dimension tables--the fact table is at the center of the
star, and the dimension tables form the points. At CMS, dimensional modeling is used for data mart
design.
Dimensional models must comply with standards in the following areas:

Diagramming conventions

Naming conventions

Relationships

Required logical metadata

Best practices (these are not standards, but compliance is expected unless the model developer can
show that specific circumstances justify an exception).
Note: The logical view of the dimensional model may not be in third normal form, but should be similar
in structure to the physical view.
2.
Diagramming and Documentation Conventions

Generate the final diagram using the standard modeling tool.

In the Model Properties dialogue, select the Dimensional feature within the Enable Modeling
Features area of the dialogue.

If capturing transformation logic in the standard data modeling tool, in the Model Properties
dialogue, select the Dimensional and Data Movement feature within the Enable Modeling Features
area of the dialogue..

Select Dimensional Modeling as the Physical Notation, and Select IDEF1X as the Logical
Notation.

Include the name of the model (or project), the name of the project contact, and the last
change date in a label within the graphic.

Include all dimension and fact tables in a single data model, but create a separate subject area for
each fact table and its associated dimension tables. This will enable the printing of a separate
graphic for each subject area when the model is reviewed.

Visual enhancements to the model (color, special fonts, etc.) are optional. If visual enhancements
are used, provide a legend within the graphic explaining the significance of each visual feature.

Separate physical and logical models (as distinct from physical and logical views within a single
model) are not required or recommended for dimensional models.

In addition to the documentation produced from the data model, a cover document is required.
This document should describe the general features of the model and any unusual characteristics.
The following should be included:

Grain statements for all fact tables and dimension tables.
The grain statement for a fact table defines the lowest level of detail needed to support the
business analytical requirements. E.g., "The grain of the Annual Hospital Bills by State
Fact Table is the year in which the medical service was rendered, the State of the patient's
residence, and the hospital identifier."
The grain statement for a dimension table identifies the natural key of the table (the natural
key may be single-part or concatenated). E.g., "The grain of the Provider dimension table
is provider number"; "The grain of the Beneficiary Demographics dimension table is age,
ethnicity, and gender."





Method of handling any changing dimensions.
Handling of semi-additive or non-additive facts (facts which usually cannot be
summarized--e.g., patient counts across multiple health care facilities cannot be
summarized to produce a count of distinct Medicare beneficiaries, and temperature
readings taken at various locations in a building would typically be averaged rather than
summarized).
Explanations of any dimensions with multiple hierarchies (e.g., a time dimension with
fiscal year and calendar year hierarchies).
Explanations of any degenerate dimensions. (Degenerate dimensions usually occur in lineitem-oriented fact table designs. For example, if an invoice number is captured in the fact
table, but all the other fact table items are at the level of the invoice line item, the invoice
number might be considered the "key" of a "degenerate"--i.e., missing--invoice
dimension.)
Explanations of any "factless fact tables." (A "factless fact table" may be used to count
occurrences of an event where no information is needed except the count itself. It may also
be used to record events that did not take place--e.g., no sales occurred for a given
product.)
3.
Naming Conventions
Business Names: General Rules

If a data object in the dimensional model corresponds exactly to one in the source system or in
another dimensional model approved by DA, use the name from the source system or other
model. Exception: If the object exists in the EDM, and the EDM name is different from the
source or other model name, the EDM name is preferred. Consult DA if there is an overriding
reason for not using the EDM name in this situation.

Spell out each word in any name that is added; do not use abbreviations. Use of acronyms in
names is limited to those acronyms on the Standard Terms and Abbreviations list.

Make nouns singular, except in cases where the plural is the only form commonly used--e.g.,
"savings," referring to an amount saved.

Use only the alphabetic characters A-Z and the space character. Do not use punctuation marks or
special characters, including the slash (/) or the hyphen (-). Numbers are not allowed.

Do not use possessive nouns--e.g., use "Beneficiary Health Assessment" rather than "Beneficiary's
Health Assessment."

Use a maximum of 120 characters, including spaces.
Business Names: Fact and Dimension Tables (Entities)
Note: Although most discussions of dimensional modeling refer to fact and dimension "tables" and make
no distinction between physical and logical terminology, ERwin uses the term "entity" in the logical
view.

The name must be a noun (e.g., "Beneficiary") or noun phrase (e.g., "Beneficiary Screening").

If DA has provided "conformed" dimension tables, use the existing business or logical names for
these tables.

Use "Fact" as the last word in a fact table name (but do not use "Dimension" as the last word in a
dimension table name).

Create fact table names that identify the business process or entity on which information is being
collected--e.g., "Medical Insurance Enrollment Fact," identifying a process, or "Hospice Usage
Fact," identifying an entity (in this case a concept).

Include a modifier in each fact table name that indicates the grain of the table--e.g., "Claim
Summary Month Fact."
Business Names: Elements (Attributes)

Each element or attribute name must include a representation term (class word) and an object class
term (prime word), optionally, a property and qualifier term (modifier). The representation term
(class word) must be the last word in the name. For more information on naming elements, see the
Attribute standard.

Make each element or attribute name unique within the data model. This is a general requirement,
but there may be exception situations. If a waiver is justified, see the DA web page for the
Standards Waiver Request Form and instructions for its submission.

If a "conformed" fact is being reused, and the existing level of granularity is matched, use the
existing name.

If aggregate or summary data from a source system is used, the data name must indicate a total or
summary--e.g., "Travel Total Amount."

For an element or attribute that captures a quantity, include the unit of measure in the element
name--e.g., Claim Blood Deductible Pint Quantity.

If the element or attribute captures a count of unique or non-duplicated values (e.g., a count of
unique beneficiaries in a set of Medicare claims), the element name must include the word
"unique" (e.g., "Beneficiary Unique Count").
Note: Consult DA's list of standard business names, acronyms, and abbreviations while
developing business names. The Standard Name and Abbreviation List will list approved terms,
abbreviations, and acronyms. The Standard Terms and Abbreviations List is available from the
Standards Terms page (accessible from the Data Administration home web page).

If a needed term is not on the approved list, follow the procedure to request on the Standard Terms
page to request a new term, which is accessible from the Data Administration web page.
Physical Names
Business or logical names must be converted to physical names using a standard method and standard
abbreviations, and physical names must be approved by Data Administration. Detailed standards and
procedures for the generation of physical names may be found in topic: Assign a Column or Element
Name.
4.
Relationships
If a fact table has multiple relationships to a single dimension table, then each relationship must be
assigned a unique rolename. Other relationships need not be named.
5.Required Logical Metadata
Table and Element Descriptions

The description must be a noun phrase. If necessary, add one or more sentences to complete the
description. E.g., the element "Provider Payment Scheduled Date" might be defined as "The
scheduled date for payment to the provider" (noun phrase). "This date is considered to be the date
of payment, since no information on the actual payment date is available" (additional sentence).

The description must be clear, precise, and unambiguous. It must identify the data object and
distinguish it from all other objects. Add examples or exclusions as needed to improve clarity.

If the element is in a time dimension, specify how the unit of time is being measured. E.g., does
"year" refer to calendar year or fiscal year; is "month" equivalent to calendar month, etc.

Describe each data object in terms of business rules, not in terms of information processing rules
or physical considerations.

If a dimension table or element in the model corresponds exactly to one in the source system or in
another dimensional model approved by DA, use the description from the source system or other
model. Exception: If the item exists in the EDM, and the EDM description is different from the
description on the source system or other model, the EDM description is preferred. Consult DA if
there is an overriding reason for using the source system description in this situation.

If a description from the EDM or a source system model is being reused, it is not necessary to
copy the source description into the project model. Instead, reference the source using the
following format:
REFER TO: xxx, where xxx is the model name, exactly as in the source model.
Note: Cross referencing is permitted only if the name of the data object is exactly the same as the
name in the source model.
Elements: Additional Metadata Requirements

Domain. Specify the valid values or range of allowable choices.

If the domain is a set of code values, list the valid values and provide a description for each
value. This information can be entered in the standard modeling tool (in the physical view
of the project model), documented in a spreadsheet, or provided in another appropriate
format. Exception: If the number of codes is large and is documented elsewhere, provide a
reference to the source document location. Enter this information at the end of the element
description. Begin a new paragraph and use the following format:
VALID VALUE SOURCE: xxx, where xxx is free-form text.


If the domain is a range of values (e.g., numeric, 1-33), document the range in the physical
view of the project model. Only a single range can be entered in the modeling tool. If
there are multiple value ranges, document them in an appropriate format.
Format. Identify the element data type, such as numeric, character, date, etc. If non-integer values
are needed, state the precision (the number of decimal places to be accommodated). If very large
numbers are expected, over a billion, state the largest value expected. This will help the DBA
select the correct data type.

Security Classification. If this element captures sensitive data, specify the degree of sensitivity.
Enter this information at the end of the element definition. Begin a new paragraph and use the
following format:
SECURITY CLASSIFICATION: xxx, where xxx is free-form text.

Alternate Business Names. If alternate names are in use for the same element within the
business community, enter the alternate names in Erwin’s User Defined Properties. Start a new
paragraph and use the following format:
ALTERNATE BUSINESS NAME(S): xxx, where xxx is free-form text.

Source System. If the source for the element data is a CMS system or database, enter the acronym
for the source system at the end of the element definition. Start a new paragraph and use the
following format:
SOURCE SYSTEM: xxx, where xxx is the system acronym.
6.Best Practices
Fact Tables

For every business process there should be at least one fact table.

Each item in a fact table should have a default aggregation (or derivation) rule--e.g., sum, min,
max, semi-additive, not additive. Any complexities in the aggregation method must be
documented.
Enter the aggregation/derivation information at the end of the definition. Begin a new paragraph
and use the following format:
DERIVATION RULES: xxx, where xxx is free-form text.

The grain, or granularity, of the fact table should be at the lowest level for which a need has been
identified and a requirement approved. Performance and storage constraints must also be
considered.

The grain of all items in the fact table should be the same. If there is a need for aggregation at
more than one level, a separate fact table for each level of aggregation may be needed.
Note: Aggregation tables are either (a) transparent to the user, such that all SQL is written to go
against the lowest level of granularity, or (b) explicit--i.e., seen and queried by the user. Since the
dimensional model is, among other things, a tool for communication with the user, it is preferable
to include only those tables that the user will see.
Dimension Tables

Each dimension table has one and only one lowest level element, called the dimension grain.

Dimension tables that are referenced or are likely to be referenced by multiple fact tables are
"conformed dimensions." If conformed dimensions already exist for any of the dimensions in the
model, their reuse is expected. If new dimensions with potential for usage across the agency are
being developed, the design must support anticipated cross-agency needs.

Each non-key element should appear in only one dimension table.

Most models should have at least one period or time dimension. There may be more than one
period dimension. Date and time may be split into two separate dimensions, especially if time is
being captured at the hour or minute level.

If a dimension table includes a code, in most cases the code description should be included. For
example, if branch locations are identified by a branch code, and each code represents a branch
name, both the code and the name should be included. An alternative is to include the description
and omit the code--e.g., State = California, Status = Active.

Generally, there should be no more than twenty dimension tables per fact table; the designer
should provide justifications if more than twenty dimension tables are required.
Keys

The primary key of a dimension table should be a surrogate key. A source system production key
should not be used as a primary key.

The primary keys of the dimension tables should be included in the fact table as foreign keys.
Together these (and only these) foreign keys make up the fact table primary key (in the logical
view).
1.6. Appendix: Summary of Related DA Standards, Forms, and Tools
Standards




Dimensional Model
Standards
Entities
Attributes
Assign a Column or
Element Name.
Forms


Tools
Data Services Request
Form
Data Standard Waiver
Request Form
Access forms from the main Data
Administration web page.

DA Standard Tools
Access information about DA tools from
the main Data Administration web page.